/AtrialFibClassifierHDA

Final Project developed for Human Data Analytics (HDA) course during MSc in Data Science at University of Padova (UNIPD)

Primary LanguagePython

AFib Classifier from ECG signals using Spectrogram and Feature Based Approaches

The project consists of a different multiclassifier built following two approaches. The first one, the spectrogram approach, transforms signals into a spectrogram and predicts its class like it was an image. The second one relies on features of Ecgs.

How to run

Feature based approaches

  • Feature signal preprocessing: python AFib.py features preprocessing

  • Feature RNN training with default epochs: python AFib.py features train rnn

  • Feature RNN training with given epochs: python AFib.py features train rnn 30

  • Feature RNN evaluation with default weights: python AFib.py features evaluate rnn

  • Feature RNN evaluation with default weights: python AFib.py features evaluate rnn ./weights.h5

  • Feature CRNN training with default epochs: python AFib.py features train crnn

  • Feature CRNN training with given epochs: python AFib.py features train crnn 30

  • Feature CRNN evaluation with default weights: python AFib.py features evaluate crnn

  • Feature CRNN evaluation with given weights: python AFib.py features evaluate crnn ./weights.h5

Spectrogram based approaches

  • Spectrogram signal preprocessing: python AFib.py spectrogram preprocessing

  • Spectrogram CNN training with default epochs: python AFib.py spectrogram train cnn

  • Spectrogram CNN training with given epochs: python AFib.py spectrogram train cnn 30

  • Spectrogram CNN evaluation with default weights: python AFib.py spectrogram evaluate cnn

  • Spectrogram CNN evaluation with default weights: python AFib.py spectrogram evaluate cnn ./weights.h5

  • Spectrogram CRNN training with default epochs: python AFib.py spectrogram train crnn

  • Spectrogram CRNN training with given epochs: python AFib.py spectrogram train crnn 30

  • Spectrogram CRNN evaluation with default weights: python AFib.py spectrogram evaluate crnn

  • Spectrogram CRNN evaluation with given weights: python AFib.py spectrogram evaluate crnn ./weights.h5